What you are looking at is a visual representation of words used on the Twitter stream and how frequently they are used by people. The large slow words have the highest frequency of use. The small fast words have the least. You can hover over a word and see some information about it. The dataset of words is rolled over on an hourly basis so that it is possible to identify trending words or topics faster.
I guess this is sort of like a mashup of tag cloud, Snellen eye charts, and some crazy web technology from that 90's that was once called a marquee. For cats.
This was written using PHP, Redis, and Twitter's fabulous Boostrap CSS Framework. This is still in a sort of "beta" so if you have any suggestions, questions or comments you can reach me on twitter.